Abstract
Currently, the application of deep learning to solving problems associated with traditional surveillance video analysis has become one of the research hot topics. The video action detection is referred to as detecting the temporal segments containing the action in the video as temporal action proposals. The existing work is mainly classified into two categories: one is to use the low-level details of video to generate action proposals; the other is to use the high-level semantics of video to generate action proposals. By deeply researching the video action detection methods based on deep learning, this paper is an attempt to find out problems with the existing methods and put forward some suggestion for improvement.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
